Neuroblastoma is a common childhood tumor, and approximately 40% of patients have aggressive metastatic disease (stage 4) when diagnosed. With improved therapy, survival has increased to 35-40%. Problem: It is not possible at diagnosis to predict which patients will be long-term progression-free survivors (PFS) and which will succumb to disease. Hypothesis: Molecular "signatures" that are derived from microarray analyses of tumor RNA and DNA will define subgroups that have either excellent or poor outcomes. We discovered a 55 gene signature from expression profiles of MYCN gene non-amplified tumors that identifies patients with 79% and 16% PFS. Overall goal: Continue developing therapeutically relevant genomic classifiers for these patients. Specific aims: 1) Determine if RNA expression profiles predict PFS. 2) Determine if DNA signatures based upon loss of heterozygosity and copy number abnormalities predict PFS and if combining DNA and RNA signatures improves accuracy of prediction. Research Design: MYCN amplified and non-amplified tumors will be analyzed as separate groups because they are clinically and biologically distinct. Tumors are available from the Children's Oncology Group and other collaborators with annotation and clinical follow-up. Signatures derived from RNA and DNA microarrays will be used to build validated molecular classifiers that predict the likelihood of PFS. RNA: Initially, expression profiling with Human Exon (HuEx) and standard HG 133 microarrays will be compared to determine if HuEx signatures have similar or better accuracy in predicting PFS. The optimal platform will be used to discover signatures with approximately 337 tumors. Clinically applicable TaqMan(r) Low Density Arrays (TLDA) will be designed for microarray signature genes and tested on the same RNAs to validate their predictive ability. Finally, an independent external set of approximately 210 tumors will be tested with TLDA assays to confirm their validity. DNA: DNA from the same specimens originally used for RNA studies along with paired normal cells will be tested with high density 500K SNP arrays to determine if DNA and DNA + RNA signatures predict PFS. If so, clinically applicable TaqMan(r) DNA assays will be designed from SNP microarray results and tested using the same DNAs. Last, the external validation set of tumors will be tested with TaqMan. DNA assays to confirm their clinical value alone or with TLDA RNA assays. Summary: These are the first and currently only studies aimed at defining subgroups among clinically defined high-risk stage 4 patients. Prediction of long-term PFS using genomic classifiers will facilitate assignment of treatment and development of more effective therapy. [unreadable] [unreadable] [unreadable]